By 2026, Microsoft, Alphabet, Meta and Amazon are expected to invest, combined, around US$650 billion in AI capacity, according to Bridgewater Associates. It is the largest commitment of corporate capital in history. Gartner projects global spending of US$2.5 trillion this year. For any C-level, the message is clear: AI is no longer a choice, it has become a pillar of business.
Being a pillar means that AI is not a tool that you take, use and keep. It influences how demand is forecast, optimizes services, understands the consumer and decides. Change the company’s architecture before it appears in the budget.
The bitter account of generative AI
In August 2025, MIT published the report The GenAI Divide: State of AI in Business 2025, which, within the sample analyzed, concluded that, of the companies that invested between US$30 billion and US$40 billion in generative AI, 95% did not obtain a measurable return on the bottom line.
Continues after advertising
The study adopts a rather alarmist view, but the central point we must note is the importance of paying attention from the beginning to how each business can capture value from investing in AI within its operation.
Most of the time, most budgets are allocated without purpose, in isolated pilots that do not connect with central processes. Result: burned capital.
Another phenomenon helps to explain the waste: “agent washing”. The term describes the practice of rebranding chatbots, simple assistants and RPA robots as “AI agents” to justify inflated prices in the wake of hype.
Continues after advertising
FOMO is not a strategy
Most Brazilian companies are falling into the same trap. Invest in AI powered by fear of missing out (FOMO, the “fear of missing out”), create an “AI committee”, distribute ChatGPT Enterprise to everyone and wait for something magical to happen. It doesn’t happen.
Before purchasing any “AI solution”, the ruler suggested by Gartner helps you decide the path. There are three categories with very different solutions and costs:
- Assistant (LLM), for searching and synthesizing information.
- Traditional automation (RPA), for repetitive and well-mapped processes. Cheaper and with less risk.
- Autonomous agent, for complex decisions, with contextual judgment and multiple chained steps.
Most companies pay astronomical prices to solve simple problems. But there is an even more expensive trap, which predates this decision: wanting to be AI First without being Data First.
Continues after advertising
No agent delivers value on a fragmented, dirty or inaccessible basis. Before hiring technology, it is necessary to invest in the infrastructure that will support it. Anyone who skips this step is not implementing AI, they are clashing with it at scale.
Where AI really delivers value
In the financial market, especially wealth management, the segment in which I have the most ownership, there is a lot of discussion about replacing professionals who are at the forefront with clients.
I am skeptical about this scenario on the visible horizon. It is unlikely that an agent will replace the part that involves empathy and financial psychologywhether in moments of adversity or in sensitive discussions such as the inheritance to be left to children.
Continues after advertising
What AI does is multiply the consultant’s capacity. Generates reports in minutes, searches portfolios for inconsistencies, synthesizes calls. More than speed, it makes possible what was not even possible before: anticipating a need before the phone rings, identifying the signal hidden in multiple emails.
Real disruption has the potential to impact the entire operation, but only those who truly enter with purpose capture value.
The path of those who are getting it right
The starting point has to be the pain, not the tool. Before any solution, map the main bottlenecks and be clear about why the agent. There are real opportunities in both the front and back-office, but as the front is more glamorous, it tends to capture most of the attention, often without a defined purpose. Without this clarity, AI becomes an expense, but with it, it becomes a growth driver.
Continues after advertising
There is also a lot of discussion about buying the solution or developing it internally.
I think this depends a lot on the company’s ambition. For those looking for timely efficiency, buying from specialized suppliers is the fastest way. For those seeking scale, developing proprietary solutions is non-negotiable: This is the only way to customize AI to customer needs and build a real competitive advantage.
Finally, defining KPIs before the pilot is essential, but the full ROI only reveals itself throughout the execution. In this game, certainty takes time, and those who are winning are those who test, learn and scale.
Perhaps the most decisive point: transformation does not happen without buy-in from strategic leaderswho need to set an example. No internal communication can replace the effect of seeing senior management using AI on a daily basis.
Enter the game the right way
The division that appears in the market is not between large and small companies, it is between those who are already testing and those who are still waiting to find out for sure. AI will be for this decade what the internet was for the 2000s: a structural reconfiguration of the economy. Those left out will pay a high price.
What separates the two groups is culture, not technology. No AI platform compensates for an organization that decides in the rearview mirror. Transformation is born from the willingness to question untouchable processes and execute with consistency.
Adopting AI is like enrolling in a gym: enrollment doesn’t deliver results, and anyone who expects transformation in three months gives up before reaping anything. The gains come to those who show up every dayadjusts load, and measures progress across quarters. The rest is noise.
